Empress Builds Base for Small-Molecule Innovation
By Wayne Koberstein, Executive Editor, Life Science Leader
Follow Me On Twitter @WayneKoberstein
You can know what could be long before you know what is. That axiom is especially applicable to the biopharma industry, where people typically make large bets on new prospects years or decades before proof confirms their worth. So, industry veterans depend on harbingers to guide their guesses about the future. One harbinger of big things to come, meaning real leaps in what we call innovation, consists of multiple advances suddenly flooding into a long-tranquil space. Case-in-point: small molecule drugs.
Empress Therapeutics is developing its own new wave of medicines in the category of small molecules, as are a growing number of other companies currently. But it is also taking a significant second step into the unknown — building an AI-based technology for discovering therapeutic chemistry among the countless native chemicals our own bodies produce.
“It’s not like small molecules ever went away,” says Jason Park, Empress CEO and cofounder. “It is still the most prescribed modality and at least 50 percent of pharma sales. It is still the only drug modality that you can consistently take orally, at home or anywhere, the only modality that gets inside of cells and readily accesses all parts of the body, and can generally be manufactured and distributed relatively easily. So, small molecules have their advantages.”
Park says he recognizes that biologic medicines solve some problems that small molecules have not, so far at least. He quickly adds that one major advantage of biomedicines stems from their being grounded in genetics, with the information encoded in DNA pointing to the specific biologic importance of those molecules as well as providing a means for rapid design and testing of potential products. He argues that the genetic code also contains instructions pointing to in vivo chemistry that can fight disease, and AI assistance is a giant advantage for deciphering these clues. “Artificial intelligence is just one tool making the discovery of small molecule chemistry a lot faster and cheaper.”
Language Evolves
Since the days before biologics swept the pharma industry, the language used to describe the action of drugs has changed dramatically. In that era, a scientific paper on the action of a particular drug in the body might typically describe its metabolic pathway, a long sequence of chemical reactions, rather than the precise molecular or cellular mechanisms likely to appear in a slide deck for a contemporary biotherapeutic. If small molecule science can follow clues held by genomes and other biological entities, it could inspire a new, even more precise language of drug action.
“In this case, we’re not only finding the targets, we’re also finding compounds that are designed to hit those targets,” says Park. Picture a gene sequence translated into a mathematical expression for a drug’s chemical formula. Serving as the master translator would be AI, augmenting human brain power with the lightning speed and dexterity of computers continuing to evolve at the speed of Moore’s Law, all while the information mined from the genome expands at the speed of light. You will soon recognize the science of small molecule drugs is not returning, but taking on an entirely new form — one that may transform all biopharma discovery and development.
“The more information we have about what’s causing a disease in humans, the better equipped scientists are to design a drug that targets it,” says Park. “Many companies fit that model — start with disease targets, ideally defined by human genetics, then design a chemical compound to drug the target. We’re doing something different; we use genetics to start with drug-like chemistry that accounts for differences between health and disease.”
So, rather than embellishing on known chemicals, Empress constructed a platform for creating new therapeutic compounds from genetics, according to Park. “AI can and is being used to model interactions between chemical compounds and disease targets. But the way that we’re using AI and machine learning is for the interpretation of genetic data. We use AI to read instructions for chemistry encoded into genetic data. In the past couple of years, the exponential growth of sequence data and computational tools like natural language processing has enabled us to better understand how DNA programs cells to synthesize chemistry.
“There’s so much information in genetic data. That’s where I think the field is going, and that’s why we started Empress. If you can use genetics to predict chemistry, you might be able to discover small molecule drugs with the speed, precision, and genetic determinism of protein, cell, and gene-based therapies.”
Faith To Confidence
How can the company be certain that its system will “reliably and predictably generate drug candidates,” as it declares? What is feeding its confidence in ultimate success? Park answers forthrightly. “When we started, it was a leap of faith. The general consensus was our mission couldn’t be done, which just encouraged us to go on and try to make it happen. To be fair, there were reasons to believe.”
He points to the long-held knowledge that chemistry inside our bodies is the result of evolution. “The number of chemical experiments conducted by Nature is on a scale quadrillions-fold beyond human capabilities. We reasoned that some of the results would be evident in the incredible information system that is DNA. We knew proteins and peptides are encoded in DNA. And we knew incredibly important substances such as neurotransmitters and steroids were proof that compounds from the body could lead directly to small molecule medicines.”
Empress gained material confidence in its mission during the past year, when the company used its platform to create a dozen new small molecule candidates for its own pipeline development — with significant results. “We’ve been putting our compounds through the same battery of tests that every drug must go through before entering the clinic, and the data just kept coming back positive. The molecules are potent and selective against important targets, they were non-toxic, they had well-behaved pharmacology in animal models,” Park says. “You don’t usually put 10 compounds into a model and have all of them look promising.”
Park describes the state of his company as poised to put the pedal to the metal. Empress took the first steps toward applying its AI findings in the labs in 2020, only three years after its founding in 2017. Since then, it has focused on industrializing its approach. Currently, it is still a Series A company, having come out of stealth only last June. Now, Park estimates it is less than two years away from putting its first drugs into the clinic. “We’ve developed programs of our own to take forward. We know our approach will work; we plan to turn the crank and do this over and over again in different diseases.”
Is this the best of times or the hardest of times to introduce AI into something like small molecule drug development? In the context of potential external obstacles such as the Inflation Reduction Act and its possible effects on investment in small molecule innovation, Park acknowledges the significant volatility and uncertainty his company and others face. Yet, he also expects to see companies like his continuing to produce extraordinary advances, bringing “value and impact for human health,” not just in small molecules, but across the board in biopharma.